2 quick notes:

1, I've done some tests lately with exoplasim runs starting with 1 meter of initial global glacial cover; it takes a bit longer for the climate to settle out, at least 50 years whereas they usually are mostly settled within 20 years otherwise, but the resulting glacial cover is a lot more accurate, as well as the global average temperature. it does make some mid-latitude areas even colder than they should be, though.

earth_glactest2 .png

2, I've seen at least some indication that T85 runs are better at reproducing the precipitation patterns in Africa and India, but I've never really got around to fully investigating it; this is a run someone on discord did for just 16 years (and I think it's from back when we used an earlier version of the topography script that was heavily land-biased)

sim.png